Skip to content
Technology How it works Breeder — Hyperion Burner — Aegis Burner — MetroVolt AI-Native Architecture Magnets Fuel cycle Safety Roadmap
Solutions AI & Data Centers Defense & Government Grid & Baseload Neutron Detection Quantum
Learn Technical Library
Proof Publications Whitepapers Technical Library Open Science & Reproducibility The Honest Gates
Company About / Mission Leadership Environment Health & Safety Investors Careers Press Contact
3D Model
AI Architecture › L3 · Twin Modeling & AI
L3 · Twin Modeling & AI

The 50-100 ms Predictive Shadow

KRONOS-CTRL runs ahead of the real plant, holding a 50-100 ms forecast so controllers act on where the machine is going, not only where it is.

THE STACK · click to jumpL7Ecosystem & StrategyL6Experience & VisualizationL5Applications & CopilotsL4OrchestrationL3Twin Modeling & AIL2Data FabricL1Control PlaneL0Foundation▲tlmctl▼L3 · TWIN MODELING & AIThe KRONOS-CTRL digital twin and its predictive shadow.1KRONOS-CTRL Twinlive plant state2GNNscoupled subsystems3PINNsphysics-constrained4Anomaly Ensemblesdrift & fault detection5MPCreceding-horizon control6Predictive Shadowruns seconds aheadMACHINE TIEState estimate descends to L1 control; alerts rise to L4 / L5.KRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORTWIN MODELING & AISHEET 05REV. 2026-08L3 · AI-NATIVE STACK
L3 · Twin Modeling & AI — its place in the stack (left, click any layer) and its internal components (right). Telemetry rises; control descends.

Running ahead of reality

The predictive shadow is KRONOS-CTRL advancing its coupled state faster than real time to hold a continuous 50-100 ms forecast of the machine. At any wall-clock instant the twin publishes both the current best estimate and a projection of the state 50-100 ms into the future under the current actuation plan. Controllers and the anomaly ensemble consume the projection, so they act on the developing situation rather than the already-past present.

Why this horizon

The 50-100 ms window is chosen to match the dynamics L3 governs: it is long enough to see a growing MHD mode, an eroding plug potential, or a developing thermal excursion in time to plan a response, and short enough that the twin's surrogates stay accurate over the rollout. The fastest events, sub-millisecond vertical instability, microsecond quench, are not the shadow's job; they belong to the L1 fast loops and the hardware failsafe. The shadow owns the millisecond-to-hundred-millisecond regime between them.

The shadow is only as good as its synchronization to the real plant: each step it is corrected toward the latest measurements so the forecast does not drift away from reality. The divergence between shadow and plant is itself monitored, a growing divergence is both a model-drift signal and an anomaly signal, since it means the machine is doing something the validated physics did not predict.

The MPC horizon is aligned to the shadow so the controller never plans further than the twin can faithfully predict; extending control ambition beyond the validated forecast would be exactly the kind of overclaim the L3 design forbids.

Content reviewed August 2026 · design-and-simulation stage